Belief-function representation of statistical audit evidence

Abstract

The belief-function representation of statistical evidence has always been a problematic issue. This problem is of particular importance to auditing because it is an application domain that fits the belief-function framework very well, but that is also confronted with the frequent use of statistical evidence as well as nonstatistical evidence. Hence, there is a need for methods to represent statistical evidence as belief functions to combine them with belief functions from other evidence types and to obtain belief in final conclusions. This paper proposes such a method designed for application in an auditing context, although it can be useful in other contexts as well. The explanation of this method, as well as the comparison of its properties with those of other existing methods is the subject of this paper. ᮊ